It works well over here in Lisbon. Yesterday, we focused on the “Welcoming Visitors” and “Getting to know my Home” tasks — today it’s time for “Catering for Granny Annie’s Comfort” and “Visit my home”. Scratchy is doing well so far.
TBM1 is about autonomously learning where the things in the apartment are located — Scratchy does this by creating a semantic map. But because of facing some issues with the trained object recognition model, TBM1 didn’t earn that much scoring yet. Also in TBM2 we encountered a few initial mistakes but than there was steady progress. For each benchmark task, it is allowed to do seven runs and the final scoring is the median of the three best ones — so there’s still room for improvement today.
For more cool pictures, follow us on instagram!